14 papers
Learn Now, Use Next, Trust Later: Prequential Test-Time Learning for LLM Agents
Tong Zhao, Reed Li, Yuyang Hu +5
Adapting large language model agents during deployment requires not only retaining past experience, but also turning new observations into timely guidance. Many test-time learning…
Beyond One-Size-Fits-All: Sample-Adaptive Strategy Routing for Vision Token Pruning in MLLMs
Haiji Liang, Pengfei Zhou, Zhenglin Wan +3
Multimodal large language models (MLLMs) process hundreds or thousands of visual tokens per image, incurring prohibitive inference costs. While existing vision token pruning method…
From Saliency to Discriminability: Rank-Preserving Visual Token Pruning for VLM Rerankers
Siyi Liu, Hanjun Yang, Chenchen Zhang +7
Large vision-language models used as listwise rerankers must jointly process visual tokens from tens of candidates per query, making token pruning essential for practical deploymen…
Beyond Polarization: The Generative Constraint of Chain-of-Thought in Pointwise Reranking
Xiaoyang Chen, Jie Liu, Haijin Liang +5
In pointwise document reranking, Chain-of-Thought models typically underperform direct scoring models. While existing diagnostics attribute this to inferior classification, score p…
RePair: Turning Retrieval Failures into Counterfactual Hard Pairs
Siyi Liu, Xiaorong Zhu, Enjun Du +6
Vision-language retrieval with CLIP-style dual encoders achieves strong cross-modal performance, yet practical accuracy often hinges on localized semantic distinctions where top-ra…
EviRank: Structured Relevance Evidence for Multimodal Image Re-ranking
Enjun Du, Siyi Liu, Zirong Chen +8
Real-world image search queries are multimodal and compositional: ``find this shirt in pink'' specifies an entity to retain, an attribute to modify, and context to ignore. Yet exis…
Training Documents Reranker with Search Rubrics for Deep Research Agent
Wenhan Liu, Yu Lu, Qiaolin Xia +8
Retrieval systems help deep research agents generate high-quality answers by providing relevant documents. However, existing retrievers typically select documents through relevance…
Omni IIE Bench: Benchmarking the Practical Capabilities of Image Editing Models
Yujia Yang, Yuanxiang Wang, Zhenyu Guan +11
While Instruction-based Image Editing (IIE) has achieved significant progress, existing benchmarks pursue task breadth via mixed evaluations. This paradigm obscures a critical fail…
Beyond Closed-Pool Video Retrieval: A Benchmark and Agent Framework for Real-World Video Search and Moment Localization
Tao Yu, Yujia Yang, Haopeng Jin +17
Traditional video retrieval benchmarks focus on matching precise descriptions to closed video pools, failing to reflect real-world searches characterized by fuzzy, multi-dimensiona…
VDE Bench: Evaluating The Capability of Image Editing Models to Modify Visual Documents
Hongzhu Yi, Yujia Yang, Yuanxiang Wang +18
Image editing models can modify visual content through natural language, but editing dense documents requires precise local changes amid tightly coupled text, tables, and layout. W…
Rank4Gen: RAG-Preference-Aligned Document Set Selection and Ranking
Yongqi Fan, Yuxiang Chu, Zhentao Xia +9
In the RAG paradigm, document ranking determines the evidence available to downstream generators. Through controlled analysis, we identify two phenomena underexplored by existing r…
FastMTP: Accelerating LLM Inference with Enhanced Multi-Token Prediction
Yuxuan Cai, Xiaozhuan Liang, Xinghua Wang +7
As large language models (LLMs) become increasingly powerful, the sequential nature of autoregressive generation creates a fundamental throughput bottleneck that limits the practic…
TFRank: Think-Free Reasoning Enables Practical Pointwise LLM Ranking
Yongqi Fan, Xiaoyang Chen, Dezhi Ye +6
Reasoning-intensive ranking models built on Large Language Models (LLMs) have made notable progress. However, existing approaches often rely on large-scale LLMs and explicit Chain-…
Best Practices for Distilling Large Language Models into BERT for Web Search Ranking
Dezhi Ye, Junwei Hu, Jiabin Fan +4
Recent studies have highlighted the significant potential of Large Language Models (LLMs) as zero-shot relevance rankers. These methods predominantly utilize prompt learning to ass…